Tài liệu ADVANDCÉ IN TAXATION VOLUME 15 pdf

Our findings using actual export data are independent of otherwise available data demonstrating a general growth in the use of DISC/FSC entities and the sales volume of these entities.. T

THE EFFECT OF EXPORT TAX INCENTIVES ON EXPORT

VOLUME: THE DISC/FSC EVIDENCE

B Anthony Billings, Gary A McGill and Mbodja Mougoué 1

IMPLICATIONS OF BENCHMARK STATE AND LOCAL TAX RATES FOR MEASURES OF ESTIMATED IMPLICIT TAXES

TAX ADMINISTRATION PROBLEMS: GAO-IDENTIFIED

SHORTCOMINGS AND IMPLICATIONS

Philip J Harmelink, Thomas M Porcano and

IMPLICIT TAXES AND PROGRESSIVITY

THE ASSOCIATION OF CAREER STAGE AND GENDER

WITH TAX ACCOUNTANTS’ WORK ATTITUDES AND

BEHAVIORS

Suzanne Luttman, Linda Mittermaier and James Rebele 111

v

THE DETERMINANTS OF STAFF ACCOUNTANTS’

SATISFACTION WITH SERVICES AT KOREAN DISTRICT

TAX OFFICES

TAX POLICY EFFECTIVENESS AS MEASURED BY

RESPONSES TO LIMITS PLACED ON THE DEDUCTION OF

EXECUTIVE COMPENSATION

Wayne State University, USA

USA

College, South Korea

vii

University of New Hampshire, USA

University, USA

EDITORIAL BOARD

EDITORThomas M Porcano

Miami University

ASSOCIATE EDITORCharles E Price

STATEMENT OF PURPOSE

Advances in Taxation (AIT) is a refereed academic tax journal published annually.

Academic articles on any aspect of federal, state, local, or international taxationwill be considered These include, but are not limited to, compliance, computerusage, education, law, planning, and policy Interdisciplinary research involving,economics, finance, or other areas is also encouraged Acceptable researchmethods include any analytical, behavioral, descriptive, legal, quantitative,survey, or theoretical approach appropriate for the project

Manuscripts should be readable, relevant, and reliable To be readable,manuscripts must be understandable and concise To be relevant, manuscripts must

be directly related to problems inherent in the system of taxation To be reliable,conclusions must follow logically from the evidence and arguments presented.Sound research design and execution are critical for empirical studies Reasonableassumptions and logical development are essential for theoretical manuscripts.AIT welcomes comments from readers

Editorial correspondence pertaining to manuscripts should be forwarded to:Professor Thomas M Porcano

THE EFFECT OF EXPORT TAX

INCENTIVES ON EXPORT VOLUME: THE DISC/FSC EVIDENCE

B Anthony Billings, Gary A McGill

and Mbodja Mougou´e

ABSTRACT

This article examines the sensitivity of U.S exports to the availability of export incentives offered under the Domestic International Sales Corporation (DISC) and the Foreign Sales Corporation (FSC) provisions of U.S tax law Evidence on the efficacy of export tax incentives is mixed The history of the DISC/FSC tax incentives provides a natural experiment to address the ques- tion of the effect of tax incentives on export volume We examine the relation

of U.S export volume to the availability of these export tax incentives from

1967 to 1998, controlling for product class and important macroeconomic variables, and find evidence of a positive association between the level of U.S exports and the existence of the export incentives offered under the DISC/FSC provisions However, this association depends on product type Our findings using actual export data are independent of otherwise available data demonstrating a general growth in the use of DISC/FSC entities and the sales volume of these entities The latter data suffer from an interpretation problem because changes in the number of special export entities used and their sales volume do not necessarily correlate with changes in actual export levels over time The approach we use in this study is an attempt to overcome

2 B ANTHONY BILLINGS, GARY A McGILL AND MBODJA MOUGOU ´E

this limitation The reported results have implications for both tax policy regarding the design of export tax incentives and the European Union’s claim that U.S export tax incentives have damaged U.S competitors in foreign trade.

1 INTRODUCTION

Research suggests that national governments can help private industry increase itsglobal market share of products under imperfectly competitive market conditions.The set of available governmental actions includes: (1) imposing tariffs onimports; (2) funding technological innovation; (3) forming export cartels; and (4)offering export incentives.1This article examines the sensitivity of U.S exports tothe availability of export incentives offered under the Domestic International SalesCorporation (DISC) and the Foreign Sales Corporation (FSC) provisions of U.S.tax law

Evidence on the efficacy of export tax incentives is mixed U.S critics argue thatsuch provisions constitute an unwarranted tax benefit to U.S exporters with little

or no real contribution to improving the U.S balance of trade position Conversely,the European Union (EU) has long argued that these tax incentives represent anillegal trade subsidy and that their existence has damaged U.S competitors inworld trade In the midst of this ongoing controversy regarding the influence ofexport tax incentives, very little research has addressed the relation of these U.S.tax incentives to actual export activity (as opposed to growth in the use of thesetax-favored export entities).2The history of the DISC/FSC tax incentives provides

a natural experiment to address the question of the effect of tax incentives onexport volume

We examine the relationship of U.S export volume to the availability of theseexport tax incentives for the time period 1967–1998, controlling for product classand important macroeconomic variables The results provide evidence on whetherU.S tax policy has assisted the U.S private sector in maintaining or increasingits export market share We find evidence of a positive association between thelevel of U.S exports and the existence of the export incentives offered under theDISC/FSC provisions However, this association depends on product type.The remainder of this article is organized as follows Section 2provides anoverview of the DISC/FSC provisions and a review of the literature that provides aframework for examining our research question We define the research variables

results inSection 5.Section 6provides sensitivity tests, andSection 7concludesthe paper

The Effect of Export Tax Incentives on Export Volume 3

2 BACKGROUND

2.1 DISC/FSC Provisions

In the late 1960s, the United States faced declining productivity for exportedstaples and a concomitant deteriorating trend in its trade balance with majortrading partners Numerous legislative proposals were introduced in the U.S.Congress to deal with this eroding trade position The DISC legislation eventuallybecame law in 1971, with the hope that this special tax benefit would stimulate theexport of U.S.-produced goods.3 The U.S trading partners almost immediatelyreacted negatively To address the complaints filed by the General Agreement

on Tariffs and Trade (GATT), the DISC was essentially abandoned in 1984 andreplaced with the FSC regime.4In 1999, the World Trade Organization (WTO),the successor to the GATT, was successful in requiring that the United Statesrepeal the FSC provisions In 2001, the U.S Congress developed a replacementfor the FSC (the “extraterritorial income exclusion” regime) only to have this newprovision successfully challenged by the EU in 2001 By the end of 2002, theUnited States faced a call to repeal the extraterritorial income exclusion regime.The DISC was essentially a U.S “paper” corporation serving as an alter ego

of its U.S.-related supplier or principal The DISC granted tax-deferral privileges

to U.S manufacturers that directly exported U.S.-produced goods The DISCincentive was intended to aid U.S exporters competing with foreign exporters thatoperated under a territorial system of taxation or a value added tax system (VAT)

As such, the DISC was structured specifically to address export incentives offered

by the U.S trading partners Congress intended that the export tax incentive wouldincrease exports of U.S.-produced products In turn, the increased exports wereexpected to produce: (1) increased employment in the affected U.S industries;(2) stabilization of the U.S dollar against foreign currencies; and (3) increasedU.S workers’ productivity.5In this regard, the cost of the DISC program in terms

of lost tax revenue was considered a small price to pay for the achievement of theaforementioned benefits

Soon after the passage of the DISC legislation, several problems arose withrespect to its provisions First, competing foreign countries argued that the DISCincentives were an illegal export incentive under Article XVI of the GATT.Second, some U.S businesses and policy makers argued that the DISC provisionsfavored large companies or specific types of industries and were merely a taxshelter Third, estimates of additional exports generated by the DISC incentivesvaried widely.6Fourth, the DISC legislation was met with disdain by EU membersthat subsequently launched a formal request to GATT calling for the UnitedStates to remedy alleged violations of GATT (Jackson, 1978) According to the

4 B ANTHONY BILLINGS, GARY A McGILL AND MBODJA MOUGOU ´Emembers, under Article XVI of GATT, a member contracting party such as theUnited States is precluded from providing a tax incentive for exports withoutgranting a similar incentive to domestic sales (Jackson, 1978).7

In 1976, responding to the aforementioned concerns, the GATT Councilformally established a panel to evaluate the DISC with respect to Article XVI ofGATT Although not yet committed to modifying the DISC on a significant basis,

in 1976 the United States altered the mechanism by which the tax benefit underthe DISC was calculated.8 However, this change appeared cosmetic to the EUmember countries and the GAAT Council Consequently, in December 1981, theGATT panel adopted an understanding, more akin to a compromise plan, statingthe following: (1) a country is not obligated to tax economic processes outside itsterritorial waters; (2) member countries are permitted to adopt measures necessary

to avoid double taxation of foreign source income; and (3) an exporting companyshould use an arms-length pricing method (World Trade Law, 1977)

The understanding essentially approved the territorial system used by EUmembers partly because the territorial system provided only indirect tax incentivesrather than the direct export subsidy provided by the DISC provisions In addition

to calling for changes in the DISC provisions, EU members called on the UnitedStates to pay between $10 and $12 billion in compensation for lost revenues due

to the DISC.9

The United States committed in October 1982 to bring the DISC in conformitywith the GATT understanding.10 In March 1983, the Reagan Administrationapproved the general outline of a proposal to replace the DISC, which died inCommittee The proposal was, as expected, a territorial type of entity involved inexporting U.S goods In this regard, the proposed FSC provisions were meant tomirror the territorial tax system used by EU members

The eventual FSC legislation required that the export entity be a foreigncorporation organized and registered under the laws of a foreign jurisdiction.11The new FSC provisions failed to satisfy EU members who believed that the FSCalso offered an illegal export subsidy under GATT Domestic critics continued toargue that, like the DISC, the FSC represented merely a tax shelter for a smallnumber of U.S companies rather than an export stimulus

The EU’s displeasure with the FSC continued, and eventually the WTO’sDispute Settlement Body issued a final report in October 1999, finding that theFSC regime violated the WTO’s Agreement on Subsidies and CountervailingMeasures The report recommended that the FSC provisions be withdrawn byOctober 1, 2000; the United States appealed the WTO’s ruling and lost (World

In response to the WTO ruling, Congress repealed the FSC provisions and as

a replacement for the FSC regime created an exclusion for certain extraterritorial

The Effect of Export Tax Incentives on Export Volume 5income (ETI).12This Act provided for the exclusion of certain qualifying foreigntrade income arising from qualified transactions that are conducted outside theUnited States after September 30, 2000.13

The EU challenged the ETI Act on the basis that it was substantively arepackaged version of the FSC rules, which were ruled as being in violation ofinternational trade laws On August 20, 2001, the Dispute Settlement Panel ofthe World Trade Organization ruled in favor of the EU claim based on two keyissues: (1) that the ETI regime constitutes an export-contingent subsidy; and (2)that the ETI provisions do not grant the same tax benefits to U.S sales as granted

to foreign sales The United States appealed the decision, but the WTO AppellateBody upheld EU claims that the ETI is an illegal export subsidy

2.2 Research Framework

Research has demonstrated that export volume (both product amount and dollarvalue) and relative market share for exports are sensitive to price changes arisingboth from fluctuations in exchange rates and real price differences (Collie,

evidence that federal tax law changes affecting the cost of capital influenceproduct price and hence export volume (Campbell et al., 1987; Dutton, 1990;

A number of studies have examined trade flows among nations within thepurview of the product life cycle model and the Hechscher-Ohlin international trademodel (Bilkey, 1982; Karlsson, 1988; Stadler, 1991; U.S Department of Treasury,

1978) Under the product life cycle model, as a product’s life cycle matures,patents expire and duplication/substitution product production becomes possible

in other countries As a result, competitors in foreign countries may be able

to begin production of the product, resulting in stiff price competition for theproduct (Dollar, 1987; Green & Lutz, 1978; Vernon, 1966) The Hechscher-Ohlininternational trade model is broader in scope and identifies economic variablesassociated with imports and exports for products in individual countries Eco-nomic variables such as: (1) capital to labor ratio; (2) product life cycle; (3) R&Dintensity; (4) level of market imperfection; and (5) economies of scale on exportstraditionally are used to explain inter-country differences in exports and imports

The empirical work based on these models suggests that the existence of marketimperfections, technological sophistication, existence of a patent for technology,capital labor ratio, real price differences, and exchange rate fluctuations aresignificant determinants of trade flows among nations The primary mechanism

6 B ANTHONY BILLINGS, GARY A McGILL AND MBODJA MOUGOU ´Ethrough which these factors operate is product price Accordingly, to the extentthat the tax incentives offered under the DISC/FSC provisions alter producerprices, we expect to see changes in export volume of U.S products.

Prior research on the economic effects of DISC/FSC provisions has primarilybeen limited to questionnaire studies administered to users of the DISC/FSCvehicles (Bello & Williamson, 1985; Bilkey, 1982) or analytical analysis ofvarious aspects of the FSC provisions (Jacobs & Larkins, 1998) Several studieshave considered the sensitivity of export volume and market share to price changesand to tax law changes (Dutton, 1990; Feenstra, 1986) Other related work hasinvestigated the specific economic factors associated with relative market share ofglobal sales for specific product categories (Brander & Spencer, 1983a; Dutton,1990; Kim & Lyn, 1987; Laussel, 1992; Lee & Stone, 1994; Nolle, 1991; Rabino,

The studies reviewed here focus primarily on the sensitivity of export andinternational markets to price differences arising from exchange rate fluctuationsand real price differences These studies are relevant in the examination of theeffectiveness of the DISC/FSC incentives because the DISC/FSC incentivestheoretically affect the price of exports and thus the volume of exports

situations where the monopoly power of the exporter is constrained for sales tosome countries and not to other countries Dutton concluded that export subsidieswill occur most often where monopoly power is constrained on sales to countrieswith a low elasticity of import demand and monopoly power is unconstrained toother countries with a high elasticity of import demand A low elasticity of exportsupply also is shown to warrant export subsidies Dutton argues for targetedexport subsidies on goods that would not otherwise be exported

Likewise,Itoh and Kiyono (1987)investigated the effects of export subsidies ongoods that would not otherwise be exported They conclude that export subsidies

on marginal goods increase the output of such goods and decrease the output ofnon-marginal goods More precisely, export subsidies are effective only on goodsthat are normally exported in small quantities or not at all

of 1986 (TRA, 1986)14 on international trade in disaggregated industries based

on the assumption that cost differences resulting from tax law changes werepassed on to consumers through price changes He used a model developed bythe U.S International Trade Commission to assess the economic impact of priorsignificant tax changes and found that the tax law changes were reflected in thecost of capital and, therefore, affected the ultimate price of the staple.15

and exports in various business units for a number of industries For this purpose,

The Effect of Export Tax Incentives on Export Volume 7

he used the Hechscher-Ohlin International Trade Model to identify the probableeconomic variables associated with imports and exports of each country Heregressed: (1) capital to labor ratio; (2) product life cycle; (3) R&D intensity;(4) level of market imperfection; and (5) economies of scale on exports andreported that the most significant determinants of exports were the level of marketimperfection and R&D intensity

As an alternative to the Hechscher-Ohlin Model,Wells (1969)examined U.S.exports of consumer durables to ascertain if certain economic patterns could beidentified within the purview of the Product Life Cycle model Wells concludedthat the income elasticity of the consumers of the product, the ability to achieveeconomies of scale, and the cost of transportation were significant factors indetermining the duration of the cycles and, therefore, global market share of theinitial producer of the product A related conclusion was that the sophistication

of the applicable technology also determines the duration of the cycles

In summary, the reviewed studies provide strong evidence that both internationalmarket share and export volume are sensitive to price changes The influence ofexport subsidies is shown to be most effective on goods that would not otherwise

be exported without the export subsidy The results of prior research suggest thatexport tax incentives represent a viable way in which U.S companies can remaincompetitive in foreign markets as the life-cycle of a particular product matures.The reviewed studies identified the following factors as significant determinants

of export volume: (1) size of exporter (Czinkota & Johnston, 1985); (2) R&Dintensity (Mansfield et al., 1979; Schneeweis, 1985); (3) capital intensity (Koo

long-term profitability (Goldstein & Mohsin, 1987; Rosson & Ford, 1982); (6)export financing (Schneeweis, 1985); (7) stage in product life cycle (Hartzok,

affect-ing the cost of capital (Rousslang, 1987); (9) the level of market imperfection

3 VARIABLES

This study addresses the effect of the DISC/FSC tax regimes on exports byregressing quarterly export volume data (aggregate and separately by producttype) on various versions of the DISC/FSC tax regimes while controlling for con-current variation in important macroeconomic variables We examine the periodbeginning with the first quarter of 1967 and ending with the third quarter of 1998.The dependent variable (EXPORT), the volume of export product in U.S.dollars, represents the level of product exported from the United States.16Export

8 B ANTHONY BILLINGS, GARY A McGILL AND MBODJA MOUGOU ´Evolume was restated for each quarter to reflect export volume at 1982 prices toallow examination of the actual changes in exported product independent of pricelevel changes Quarterly export data were obtained from the U.S Department of

Commerce, Bureau of Economic Analysis, Survey of Current Business.17

The explanatory variables are categorized as macroeconomic and tax regimevariables With respect to the macroeconomic variables, a number of the keyvariables identified by the reviewed studies are included in the estimation toaccount for variation in export volume unrelated to changes in the DISC/FSCregimes (Bernard & Jensen, 1998; Chowdhury, 1993; Schneeweis, 1985).Based on earlier studies (Lee & Stone, 1994; Mansfield et al., 1979; Schneeweis,

to be positively related to export volume Firms with greater R&D intensity,ceteris paribus, are expected to be able to better exploit the value of the resultingproperty created by the R&D in foreign markets because of less competition fortheir products (duplication or substitution of products are less possible in othercountries) R&D intensity is measured as the proportion of sales that is spent onR&D (annual R&D expenditures divided by annual sales) Yearly R&D intensitymeasures obtained from the National Science Foundation [various issues]

were matched with each of the product categories used in the study QuarterlyR&D intensity measures then are generated from these annual R&D intensitymeasures.18

The reviewed studies also have shown that U.S exports are affected byexchange rate fluctuations As products become more costly in U.S dollars due

to exchange rate fluctuations, exports are expected to decline, ceteris paribus Weuse the weighted-average exchange value of the U.S dollar against the currencies

of other G-10 Countries (EXCHANGE) to control for variation in exchange ratesover the study period (Bernard & Jensen, 1998; Chowdhury, 1993; Schneeweis,

1985) EXCHANGE for each quarter was obtained from the Board of Governors

of the Federal Reserve System, Foreign Exchange Rates, G.5 (405).

Other macroeconomic variables with the potential to affect export level include:(1) the U.S consumer price index (CPI); (2) the S&P price index (SPI); (3) the U.S.bond rate (USBOND); and (4) the U.S industrial production index (INDPROD).Although the export data used in the estimations are measured in constant dollars,changes in CPI could separately affect demand for exports For example, an in-crease in CPI could curtail domestic demand and put pressure on firms to increaseexports Quarterly Consumer Price Index-Urban data were obtained from the U.S

Department of Labor, Bureau of Labor Statistics, The Consumer Price Index The

SPI and USBOND variables help control for economy-wide effects that may beassociated with firm activity, including export levels Data for SPI were based oneach quarter’s Common Stock Price Index for the S&P 500 and were obtained from

The Effect of Export Tax Incentives on Export Volume 9

Standard & Poor’s Corporation, The Outlook Data for USBOND were obtained

from the U.S Department of the Treasury, Treasury Bulletin In periods with

increased INDPROD, exports are expected to increase because either firms pursuegreater export sales because of the increased production or greater export demandhelps fuel increased production Quarterly data for INDPROD were obtained from

the Board of Governors of the Federal Reserve System, Industrial Production,

Statistical Release G12.3.Table 1provides a summary of all the variables used

in the estimations

The DISC/FSC tax regime variables identify the particular export tax incentivestructure in place during the observation quarter These periods are identifiedbelow:

(1) Pre-1972 Prior to 1972, no special export tax incentives existed in U.S tax

laws This period allows examination of the relationship between export ume and the non-tax variables absent the tax incentives

vol-(2) 1972–1976 The initial DISC rules were in place for years after 1971 The

1972–1976 regime is expected to have a positive association with U.S exportlevel if the DISC provisions were effective in increasing exports

(3) 1977–1984 The Tax Reform Act of 1976 placed a limit on DISC benefits

for entities with adjusted taxable income exceeding $100,000 Only taxableincome attributable to export gross receipts exceeding 67% of a four-year baseperiod average was subject to deferral treatment This 1977–1984 regime isexpected to have a negative association with U.S export level relative to theprevious period because the DISC benefits were restricted

(4) Post-1984 The DISC provisions were repealed generally and replaced with the

FSC provisions in 1984 This post-1984 regime is expected to have a positiveassociation with U.S export level

To examine the effects of the DISC/FSC programs on exports on an overall basis,

we use aggregated export data SeeFig 1for a summary of exports over the ple period However, disaggregation of exports into various commodity classesenables us to ascertain whether the functional relations between export volumeand the various explanatory variables differ by product type Eleven productgroupings were selected for analysis based on data availability.19 The selectedproduct groupings, listed inTable 2, cover a large number of sectors, ranging fromnondurable goods to capital goods Seven of the product classes are unique, withtwo of the classes (consumer goods and industrial supplies and materials) furthersubdivided into two subcategories each Quarterly export data for the variousproduct categories were obtained from the U.S Department of Commerce,

sam-Bureau of Economic Analysis, Survey of Current Business (U.S Department of

10 B ANTHONY BILLINGS, GARY A McGILL AND MBODJA MOUGOU ´E

Table 1. Model Variables

EXPORT Log of quarterly export level in constant

1982 dollars (millions)

U.S Department of Commerce, Bureau of

Economic Analysis: Survey of Current

Business

R&D Log of ratio of annual R&D outlays to

annual sales aggregated by industry

Research and Development in Industry,

Surveys of Science Resources Series,

National Science Foundation (various years)

EXCHANGE Log of weighted-average exchange value

of U.S dollar against currencies of other

G-10 countries

Board of Governors of the Federal Reserve

System: Foreign Exchange Rates, G.5

(405)

CPI Log of Consumer Price Index for all

Urban Consumers

U.S Department of Labor, Bureau of

Labor Statistics: The Consumer Price

Board of Governors of the Federal Reserve

System: Industrial Production, Statistical

Release G12.3

TREND Dummy variable counter for time period 1

to N (1–127 quarters); measures slope of

export series prior to DISC introduction

Created

D1L Dichotomous dummy variable scored 0

before DISC rules effective and 1

afterward (January 1971); captures change

in the level of exports

Created

D1S Dummy variable counter scored 0 before

DISC rules effective and 1 to N for periods

afterward (January 1971); captures change

in slope

Created

D2L Dichotomous dummy variable scored 0

before Tax Reform Act of 1976 change to

DISC rules and 1 afterward (January

1977); captures change in the level of

exports

Created

D2S Dummy variable counter scored 0 before

TRA 1976 changes and 1 to N for periods

afterward (January 1977); captures change

in slope

Created

D3L Dichotomous dummy variable scored 0

before change to FSC and 1 afterward

(January 1985); captures change in level

Created

D3S Dummy variable counter scored 0 before

change to FSC and 1 to N for periods

afterward (January 1985); captures change

in slope

Created

a

The Effect of Export Tax Incentives on Export Volume 11

Fig 1 Level of U.S Exports.

Table 2. Product Groupings Used in Analysis.a

Model Product Groupings

1 Automotive vehicles, engines, and parts

2 Capital goods, except automotive

3 Civilian aircraft, engines, and parts

4 Computers, peripherals, and parts

5 Consumer Goods, except automotive (aggregate of product classes 6 and 7)

6 Durable goods (sub-class under consumer goods, except automotive)

7 Nondurable goods (sub-class under consumer goods, except automotive)

8 Industrial supplies and materials (aggregate of product classes 9 and 10)

9 Durable Goods (sub-class under industrial supplies and materials)

10 Nondurable Goods (sub-class under industrial supplies and materials)

11 Foods, Feeds, and Beverages

a Product categories were obtained from the U.S Department of Commerce, Bureau of Economic

Analysis, Survey of Current Business.

4 METHOD

The relationship between exports and the explanatory variables is estimated forthe aggregate data and for each of the product class groupings using the followingmodel in which exports and all of the macroeconomic control variables areexpressed in logarithmic form:20

EXPORTt = ␤0+ ␤1TREND+ ␤2D1L+ ␤3D1S+ ␤4D2L+ ␤5D2S

+ ␤6D3L+ ␤7D3S+ ␤8R&Dt+ ␤9EXCHANGEt

+ ␤ INDPRODt+ ␤ CPIt+ ␤ SPIt+ ␤ USBONDt+ ␮

12 B ANTHONY BILLINGS, GARY A McGILL AND MBODJA MOUGOU ´Ewhere ␮ = ␧ − ␣1␮t−1− · · · − ␣p␮t −p; ␧ is normally and independently

distributed with a mean of 0 and a variance of s2; and p is the order of the

autoregressive process to be fit.21

TREND is a simple time trend (1–127) used to capture the slope of the exportseries The DiLs and DiSs are indicator variables used to capture the impact ofeach tax regime on exports That is, D1L is a dummy variable scored 0 beforethe 1971 DISC rules became effective and 1 afterward; D1L captures the change

in the level of exports due to the initial adoption of the DISC D1S is a dummy

variable counter scored 0 before DISC rules became effective and 1, 2, 3, ., N for periods afterward; D1S captures change in the slope or growth rate of exports after

introduction of the DISC D2L is a dummy variable scored 0 before Tax ReformAct (TRA) of 1976 change to the DISC rules and 1 afterward; D2L captures the in-

cremental change in the level of exports caused by the 1976 TRA D2S is a dummy

variable counter scored 0 before TRA 1976 changes and 1, 2, 3, ., N for periods afterward; D2S captures the change in the slope or the growth rate of exports

attributable to the 1976 TRA D3L is a dummy variable scored 0 before the 1984

change to FSC and 1 afterward; D3L captures the change in the level of exports

brought about by the 1984 change to the FSC regime D3S is a dummy variablecounter scored 0 before the change to FSC and 1, 2, 3, ., N for periods afterward D3S captures the change in the slope or the growth rate of exports attributable to the

1984 change

The estimation model isLewis-Beck’s (1986)“interrupted” time series model.The model’s most appealing feature is that it allows for the assessment of eachregime’s effect on both the level and growth rate of exports With the introduction

of each new regime, the dummy variables related to that regime capture anychanges in the intercept (DL) or slope (DS) Significant parameter estimates for thetax regime variables provide evidence of an incremental tax effect after controllingfor other macroeconomic variables This approach is equivalent to fitting separateregression models for each of the tax regime periods and then comparing theresulting intercept and slope parameters (Lewis-Beck, 1986) Because the depen-dent variable is the logarithm of export level, the estimated parameters (×100) ofthe dummy variables can be interpreted as percentage changes in export level.Econometric analysis of time series data must include unit root testing becausethe validity of the empirical relation between variables is predicated on therequirement that the classical stationarity assumptions are satisfied.Granger and

the time series involved are nonstationary.22Consequently, before any attempt tomeasure the impact of export-related tax incentives on export volume is made wefirst examined the dependent and independent variables to determine whether theysatisfied stationarity conditions Two stationarity testing procedures were used

The Effect of Export Tax Incentives on Export Volume 13

fol-lowing regression equation:

Y t = ␣ + ∃(t − N/2) + ␳Y t−1+ ␧t , t = 1, 2, , N

where Y t denotes the series being tested for a unit root, (t − N/2) is a time trend, and N is the sample size.

The PP equation is employed to test three null hypotheses:

(1) H10:␳ = 1, the series Y tcontains a unit root with a drift and a time trend.(2) H20: ␣ = 0, ␳ = 1, the series Y t contains a unit root with a time trend andwithout a drift

(3) H30:␣ = 0, ␤ = 0, ␳ = 1, the series Y tcontains a unit root without a time trendand without a drift

The 5 and 1% critical values for testing these three hypotheses are taken fromFuller

in Panel A ofTable 3 All the reported statistics are significant at the 1% level,indicating that all the series are stationary and, therefore, need not be differenced

in empirical investigation

The PP test has come under attack on the grounds that its failure to reject theunit root hypothesis may be attributable to its low power against weakly stationaryalternatives.Kwiatkowski, Phillips, Schmidt and Shin (1992)(KPSS) recommend

a test of the null hypothesis of stationarity against the alternative of a unit root.23The KPSS test statistics are given by

i s are the residuals obtained by regressing the series being tested

on a constant without a trend and on a constant and a time trend, respectively Ifthe test statistic exceeds the critical values, the null hypothesis of stationarity isrejected in favor of the unit root alternative

14 B ANTHONY BILLINGS, GARY A McGILL AND MBODJA MOUGOU ´E

Table 3 Unit Root Results for the Log of Aggregated Data.

Variable Panel A: Phillips-Perron Testa Panel B: KPSS Testb

(Null Hypothesis) (Null Hypothesis)

H1: H2: ␣ = 0, H3: ␣ = 0, H1: Series is H2: Series is

␳= 1 ␤= 1 ␤ = 0, ␳= 1 Level-Stationary Trend-Stationary EXPORT −25.763 ∗ 62.109∗ 49.534∗ 0.275 0.034R&D −17.553 ∗ 56.070∗ 40.764∗ 0.333 0.107EXCHANGE −22.645 ∗ 43.007∗ 39.985∗ 0.234 0.067INDPROD −31.004 ∗ 21.037∗ 17.963∗ 0.291 0.113

H2: 4.75 and 6.22; H3 : 6.34 and 8.43 The truncation lag is twelve for the reported results, but the

conclusions are the same for other truncation lag values An asterisk ( ∗) indicates significance at the1% level.

b The test statistics for the null hypotheses of level-stationary series ˆ ␩ u and trend-stationary series ˆ ␩t

are given as follows,

e t e t −s The e i s and e∗

i s are the residuals obtained by regressing the series being tested on a constant without

a trend and on a constant and a time trend, respectively The 5 and 1% critical values are 0.463 and 0.739 and 0.216 and 0.146 for ˆ ␩ u and ˆ ␩t, respectively ( Kwiatkowski et al., 1992 , p 166, Table 1 ) The

reported test statistics are computed using lag length L that equals 12.

The results of the KPSS test are given in Panel B ofTable 3 The ˆ␩ustatistictests the null hypothesis of level stationary series, whereas the ˆ␩t statistic teststhe null of trend-stationary series Both test statistics fail to achieve statisticalsignificance at any conventionally accepted level for the dependent variable andall the macroeconomic control variables This finding indicates a rejection ofthe unit root hypothesis and is in agreement with the PP test in supporting thestationarity of all the macroeconomic variables over the 1967–1998 period

The Effect of Export Tax Incentives on Export Volume 15

5 RESULTS

variables, controlling for macroeconomic variables, for models using aggregateU.S product exports and separately for each of the eleven separate productcategories

In the model for all product exports (ALL), four of the six macroeconomic

control variables are significant in the theoretically expected direction The

R&D coefficient (0.1671) is positive and statistically significant ( p = 0.0035).

This finding suggests that R&D does exert a strong influence on export leveland is consistent with earlier studies (e.g.Schneeweis, 1985) As expected, therelation between the level of export and exchange rate (EXCHANGE) is negative

and statistically significant ( p = 0.0000) This finding is consistent with prior

research (Bernard & Jensen, 1998; Chowdhury, 1993; Schneeweis, 1985) TheEXCHANGE result implies that as the weighted-average exchange rate increases(i.e as the U.S dollar appreciates) the export levels decline Both increasedindustrial production (INDPROD) and CPI are positively related to export level.The positive association between INDPROD and exports can be viewed asevidence that, all else constant, an increase in industrial production tends to putpressure on U.S firms to pursue foreign markets or that foreign market demandfuels increased production An increase in CPI could potentially curtail domesticdemand Any decrease in domestic demand for U.S goods puts pressure on firms

to export Finally, neither the SPI nor the USBOND variable achieves statisticalsignificance.24

The results for each of the eleven product class models are mixed, but many

of the same relations exist R&D was significant and positive in five models andnegative and significant in two models EXCHANGE was significant and negative

in nine models INDPROD was positive and at least marginally significant innine models CPI was positive and at least marginally significant in eight models.SPI was significant in only three models but with mixed signs USBOND waspositive and at least marginally significant in only two models Overall, thesemacroeconomic control variable results are consistent with relations identified inprior research (e.g.Bernard & Jensen, 1998; Nolle, 1991; Yang, 1996)

In order to interpret the tax regime parameter estimates, recall that the estimates

␤0and␤1, for example, indicate the overall level and trend of exports, respectively

We evaluate D1L and D1S to ascertain whether the level and trend changed as aresult of the adoption of the DISC provisions If the coefficient on D1L is statisti-cally different from zero, the implication is that the 1971 DISC legislation had aninfluence on the level of exports Similarly, if D1S is statistically significant, weinfer that the adoption of the 1971 legislation altered the growth rate of exports.25

INDPROD 1.4168 (0.0000) 1.1203 (0.0029) 1.3232 (0.0000) −4.4618 (0.0456) 1.3656 (0.0060) CPI 2.2552 (0.0000) −0.2641 (0.6229) 2.4636 (0.0000) 1.1996 (0.7137) 2.0815 (0.0096) SPI 0.0445 (0.3980) −0.4407 (0.0000) 0.0725 (0.2240) 0.6577 (0.1424) −0.0105 (0.9360)

USBOND 0.0294 (0.5122) −0.0920 (0.2196) 0.1081 (0.0529) 0.5354 (0.1177) 0.0972 (0.5095)

Constant 1.4965 (0.0000) 2.1255 (0.0000) 1.2481 (0.0000) 1.2398 (0.0000) 1.2767 (0.0014) TREND −0.0440 (0.1586) −0.0290 (0.1894) −0.0565 (0.2011) −0.0902 (0.0030) −0.0896 (0.0000)

D1L ( +) 0.1279 (0.1578) 0.1487 (0.0233) 0.1424 (0.2455) 0.1475 (0.1030) −0.0332 (0.4421)

D1S ( +) 0.0048 (0.8755) −0.0385 (0.0642) 0.0356 (0.4189) 0.0544 (0.0725) 0.0909 (0.0000) D2L ( −) 0.0639 (0.1381) 0.1154 (0.0379) −0.0359 (0.3405) −0.0635 (0.1350) −1.2049 (0.0000)

D2S ( −) −0.0168 (0.0014) −0.0208 (0.0007) −0.0103 (0.0274) −0.0055 (0.3291) −0.0334 (0.0091)

D3L ( +) −0.0163 (0.0032) −0.0247 (0.0000) −0.0166 (0.0000) −0.0595 (0.0223) 0.0121 (0.0029) D3S ( +) 0.0538 (0.0000) 0.0823 (0.0000) 0.0273 (0.0000) 0.0236 (0.0000) 0.0113 (0.0282) R&D 0.1135 (0.1241) 0.3884 (0.0000) −0.4463 (0.0000) 0.0041 (0.9612) 0.1752 (0.0488) EXCHANGE −0.3943 (0.0003) −0.4672 (0.0002) −0.1303 (0.00411) −0.2108 (0.0001) −0.4281 (0.0000)

c The eleven product groupings are described as follows: (1) Automotive vehicles, engines, and parts; (2) Capital goods, except automotive; (3) Civilian aircraft, engines, and parts; (4) Computers, peripherals, and parts; (5) Consumer Goods, except automotive (aggregate of product classes 6 and 7); (6) Durable goods (sub-class under consumer goods, except automotive); (7) Nondurable goods (sub-class under consumer goods, except automotive); (8) Industrial supplies and materials (aggregate of product classes

9 and 10); (9) Durable Goods (sub-class under industrial supplies and materials); (10) Nondurable Goods (sub-class under industrial supplies and materials); (11) Foods, Feeds, and Beverages.

dThe R2 relates to the structural model after transformation for any autocorrelation but does not include the variance explained related to the inclusion of the models’ past

18 B ANTHONY BILLINGS, GARY A McGILL AND MBODJA MOUGOU ´E

Examination of the ALL model inTable 4suggests that on an aggregate productbasis, the 1971 introduction of DISC affected exports only marginally The

coefficient on D1L (0.0486) is insignificant ( p = 0.4505) but the coefficient on D1S (0.0429) is marginally significant ( p = 0.0557) These findings indicate

that although the adoption of the DISC legislation did not meaningfully alter thelevel of exports, it did lead to a marginal increase in the growth rate of exports,measured in the aggregate

The restriction of DISC benefits in 1977, however, appears to have significantlyaffected both the level and growth rate of exports on an overall basis.Table 4reportsthat both D2L and D2S have significant and negative coefficients, as predicted.These results indicate that the limitation of the DISC tax benefits beginning in

1977 dampened both the level of exports (a 7.10% decrease) and the export growthrate (a 2.04% decrease) Prior to the 1977 tax regime, the model intercept term

is estimated at 1.5248 (1.4762 + 0.0486) and the model slope (growth rate over

time) is estimated at−0.0258 (−0.0687 + 0.0429) Thus, prior to the change in the

DISC benefits, the real growth rate of exports was a negative 2.58% per year Afterthe change to DISC benefits in 1977, the intercept is estimated at 1.4538 (1.5248 −

0.0710) and the slope becomes more negative at −0.0462 (−0.0258 − 0.0204).

Consequently, the 1977 DISC change appears to have lowered both the level ofexports and the export growth rate measured on an aggregate basis This negativeeffect is predicted because the 1977 changes reduced potential DISC tax benefits.Finally, the introduction of the FSC in 1985 seems to have exerted a significantimpact on both the level and growth rate of exports As indicated in Table 4,

both D3L and D3S have significant positive parameter estimates in the ALL

model, suggesting that the imposition of the FSC rules in 1985 improved thelevel of exports by 1.58% (D3L= 1.58%) and the export growth rate by 3.89%

(D3S= 3.89%) Thus, the FSC rules appeared to have reversed the negative trend

in exports generated by the restricted DISC rules introduced in 1977

Examination of the eleven product class models produces somewhat mixedevidence, suggesting that the export incentives effect varies across productcategory With regard to the imposition of the 1971 DISC rules, four productcategories (1, 3, 6, and 10) produce positive and significant level effects andone product category (4) has a significant negative level effect The remainingsix product groupings (2, 5, 7, 8, 9, and 11) exhibit no significant relationshipbetween export level and the introduction of the DISC regime

As for the growth (slope) effect (D1S), Table 4 shows that five productcategories (1, 3, 4, 9, and 10) are significant at the 1% level and two productgroups (2 and 8) achieve statistical significance at the 10% level, all with theexpected positive sign Only one product grouping (6) shows a negative slope

effect that approaches significance ( p = 0.0642).

The Effect of Export Tax Incentives on Export Volume 19The effect of the 1977 DISC change on exports is captured by the estimates ofthe D2L and D2S coefficients Two of the eleven product groups (2 and 9) havethe expected significant negative level effects (D2L) Three of the product groups(1, 3, and 6) have significant positive level coefficients The findings for the 1977change slope effect are more consistent across product classes Five of the elevenproduct categories (2, 5, 6, 7, and 9) have the expected significant negative slopeeffects (D2S) Two of the remaining product groups have significant positiveslope effects (1 and 3).

The effect of the FSC regime introduction on export level is captured by D3L

significant positive sign on the D3L coefficient Six of the remaining productcategories (2, 5, 6, 7, 8, and 11) have significant negative level effects In contract

to these mixed results on the level coefficients, the introduction of the FSC regime

is consistently positively associated with the growth rate of exports as predicted.With the single exception of the insignificant coefficient for product group 4, theD3S coefficients are all significant and positive, implying a strong export growtheffect Overall, it appears that the adoption of the FSC regime had the intendedeffect of increasing growth in U.S export levels relative to the dampening effecttriggered by the 1976 legislation placing limits on DISC benefits

The predicted level effects are mixed, with product categories 1 (automotive

vehicles, engines, and parts) and 3 (civilian aircraft, engines, and parts) strating the most consistency with the expected effect from introduction of theDISC or FSC regimes Product categories 2 (capital goods), 5 (consumer goods,except automotive), 7 (nondurable goods, subclass of category 5), 8 (industrialsupplies and materials), and 11 (foods, feeds, and beverages) do not conform to the

demon-predicted level effects from the DISC or FSC regimes Predicted growth effects are

more consistent across product categories The structure of these separate productcategory results are explored in more detail in the discussion section below

6 SENSITIVITY TESTS

In this section, we conduct additional tests to determine the robustness of ourempirical findings with respect to: (1) the measurement of the export dependentvariable; and (2) two international agreements thought to have affected U.S.exports through significant changes in the relative value of the U.S dollar

In order to determine the sensitivity of our results with respect to the ment of the export variable, we redefine U.S exports as a percentage of grossdomestic product (GDP) SeeFig 2for a summary of exports relative to GDP Anincrease in export level could come from two sources: a shift from domestic sales

measure-20 B ANTHONY BILLINGS, GARY A McGILL AND MBODJA MOUGOU ´E

Fig 2 U.S Exports Relative to GDP: 1967Q1–1998Q3.

to export sales or an increase in domestic production, with some portion of thisincrease accruing to export sales Defining exports as a percentage of GDP allows

us to avoid this interpretation problem We re-estimate our model for the aggregatedata and for each of the eleven product categories using the new export definition.The empirical results are qualitatively similar to those reported inTable 4

We also evaluate the sensitivity of our empirical findings with respect to twomajor international agreements: The Plaza Accord (September 1985) at which theG-7 nations agreed to co-ordinate policy so as to substantially reduce the value

of the U.S dollar in foreign exchange markets, and the Louvre Accord (February1987), where the G-7, having decided that the dollar had fallen enough, reversedcourse and sought to stabilize the U.S dollar within a relatively narrow targetvis-`a-vis the German mark and the Japanese yen Using indicator variables tocontrol for these events (similar to the method employed to capture the tax regimeeffects), we re-estimate the models to determine whether these internationalagreements affected U.S export levels by affecting the relative value of theU.S dollar in foreign exchange markets The empirical results are qualitativelysimilar to those in Table 4 for both the aggregated data and for the elevenproduct categories.26

The last sensitivity test we conduct examines the impact of lagged independentvariables on our overall results More specifically, we re-estimate our modelusing: (a) the levels and the lagged values of the independent variables jointly;and (b) lagged independent variables only Overall, these additional tests reveal

The Effect of Export Tax Incentives on Export Volume 21that lagged values of the independent variables are not significantly related toeither the level or the growth of U.S exports during the 1967–1998 period.

7 DISCUSSION

Our research objective is to determine whether export tax incentives demonstrateany significant association with export level after controlling for other macroe-conomic factors The aggregate analysis suggests that the DISC/FSC incentivesare significantly related to changes in the level and/or growth rate of exportsover the period studied The model estimation results seem particularly crediblebecause the tax regime variables demonstrate a positive association with exportswhen the tax regime change increases tax benefits and a negative association withexports when the tax regime change reduces tax benefits However, these aggregateresults do not hold consistently for each of the eleven product groupings examined.This result suggests that although there is some detectible relationship of exportlevel to tax incentives, substantial variability exists among product groups.The observed variability of the effects of the DISC/FSC regimes among theproduct classes is in accord with prior research that suggests export subsidiesinfluence only certain products and exports to certain countries (Dutton, 1990;

subsidies influence exports primarily on products sold to countries with a highelasticity of import demand Likewise,Dutton (1990)argues that export subsidiesare most effective on goods that would not otherwise be exported or on goodsthat are exported in small quantities

The U.S Treasury previously estimated the effect of FSC benefits on differentproducts based on supply and demand elasticities (U.S Treasury, 1993).27 Theproduct categories used in the current study are rather broad compared with theTreasury data; however, some high-level comparisons are possible In the Treasurystudy, machinery (electrical and non-electrical) and transportation equipment areexpected to be the most sensitive sectors to the availability of FSC benefits (U.S

vehicles, engines, and parts) and 3 (civilian aircraft, engines, and parts) can bemapped into the Treasury’s transportation category Product classes 2 (capitalgoods, except automotive) and 9 (durable goods) are related to the Treasury’smachinery category Consistent with the Treasury findings, product classes 1, 2,

3, and 9 exhibit the greatest number of significant parameter estimates consistentwith the hypothesis that DISC/FSC benefits influence export volume in theexpected direction With 24 possible DISC/FSC effects (4 product categories with

6 level or growth parameters), 67% (16 of 24) were significant in the predicted

22 B ANTHONY BILLINGS, GARY A McGILL AND MBODJA MOUGOU ´Edirection Compare these results with categories 4, 5, 6, 7, 8, 10, and 11 where only33% (14 of 42) of the possible DISC/FSC effects were significant in the predicteddirection At the other extreme, the Treasury found that food products wereexpected to have little sensitivity to FSC benefits Product class 11 (foods, feeds,and beverages) can be mapped into the Treasury’s food category Consistent withthe Treasury results, the findings for class 11 suggest that this product class is lesssensitive to DISC/FSC benefits Only one of the six DISC/FSC regime variables(17%) is statistically associated with export level in this category in the expecteddirection In summary, the cross-product variation in sensitivity to export incen-tives observed in this study is consistent with the Treasury’s results obtained using

a different method.28

Lastly, the macroeconomic control variables used in our study produce findingsconsistent with their use in prior research (e.g.Kim & Lyn, 1987; Laussel, 1992;

intensity, exchange rates, industrial production level, and CPI are significant inthe predicted direction for the aggregate model and across many of the separateproduct class models Product group 11 is a notable exception, with none of thecontrol variables significant in the expected direction This result is not surprisingbecause this category represents food products and export levels in this categoryare expected to have low supply and demand elasticities (U.S Treasury, 1993).The results of this study must be considered in light of several limitations.Questions addressing broad macroeconomic relationships are difficult at best,and made only more so with the introduction of multinational effects Morespecifically, the effects of the DISC/FSC tax incentives captured in this studymight be explained in alternative ways Other events during the period of studymay coincide with the tax regime periods and our robustness tests did not captureall possible events Furthermore, the measures used in this study are at a fairlybroad level of aggregation and a finer partition may produce different results.Notwithstanding these limitations, the results of this study provide evidencethat export tax incentives influence export levels in the theoretically-predictedmanner We examine the relation of U.S export volume to the availability of theseexport tax incentives over the period 1967–1998, controlling for product class andimportant macroeconomic variables, and find evidence of a positive associationbetween the level of U.S exports and the existence of the export incentives offeredunder the DISC/FSC provisions However, this association is dependent onproduct type Our findings using actual export data are independent of otherwiseavailable data demonstrating a general growth in the use of DISC/FSC entities andthe sales volume of these entities (e.g.U.S Treasury, 1993) The latter data sufferfrom an interpretation problem because changes in the number of special exportentities used and their sales volume do not necessarily correlate with changes in

The Effect of Export Tax Incentives on Export Volume 23actual export levels over time The approach we use in this paper is an attempt toovercome this limitation The reported results have implications for both tax policyregarding the design of export tax incentives and the European Union’s claim thatU.S export tax incentives have damaged U.S competitors in foreign trade.

NOTES

1 See, for example,Auquier and Caves (1979),Bello and Williamson (1985),Branderand Spencer (1983a, b),Gravelle, Hughes and Farb (1976),Green and Lutz (1978),Nolle(1991),Goldstein and Mohsin (1987),Schneeweis (1985),U.S Department of Treasury(1978), andVernon (1966)

2 Data are available from the U.S Treasury regarding the growth in the number ofDISCs or FSCs over time and the export volume reported by these entities (see, forexample,U.S Treasury, 1993) However, data on changes in the number of special exportentities and their sales volume do not necessarily correlate with changes in actual exportlevels over time For example, firms may begin using FSCs for export sales that alreadywere being made outside the FSC entities Consequently, an increase in reported FSC salesmay not correspond with an actual increase in export sales The approach we use in thisstudy is an attempt to overcome this limitation

3 Revenue Act of 1971, P.L 92–178

4 A restricted version of the DISC remained after 1984 (the so-called Interest-ChargeDISC) In response to the criticism that tax deferral on DISC profits resulted in a permanentreduction in the tax liability of DISC shareholders, IRC Section 995(f)(1) was added toimpose an interest charge on the deferral of DISC income earned after January 1, 1985

to the extent deferral reduces the tax liability of affected shareholders The taxpayer cost

of the interest charge provision is partially mitigated because the interest charge is adeductible expense

5 Hearings Before the House Ways and Means Committee, 91st Cong., 2d Sess., pt 2

at 500 (1970) and H.R Rep 658, 94th Cong., 2d Sess 264 (November 12, 1975)

6 The U.S Treasury, as part of its 1981 annual report, reported that DISC tax incentivesinduced additional exports between $7.2 and $11 billion annually (U.S Department ofTreasury, 1983) According to the Treasury report, U.S exporters deferred up to $751million in taxes under the DISC program As of 1975, DISCs also were credited withproviding roughly 230,000 U.S jobs annually But the Treasury Department’s estimatesalso reported a $1.65 billion cost of the DISC program In rebuttal to the Treasuryreport, domestic critics argued that the DISC provisions created no more than $1–3billion in additional exports at a revenue cost of approximately $1.2 billion (WorldTrade Law, 1977)

7 Indeed, the DISC had several provisions that were allegedly in violation of ArticleXVI of GATT The most blatant of these were: (1) the lack of required foreign presence;(2) the failure to charge interest on deferred tax liabilities; and (3) the failure to imposearm’s length pricing between the DISC and its related supplier/principal

8 Tax Reform Act of 1976, P.L 94–455

9 This request may have been made in part because of negative sentiment against FSCproposed changes The formal request was made on July 16, 1984 The United States

24 B ANTHONY BILLINGS, GARY A McGILL AND MBODJA MOUGOU ´Eresponded to the GATT panel by pointing out the following: (1) DISCs, in substance,complied with the policy objectives and intent of GATT; (2) DISCs actually provided

a smaller tax benefit to U.S exporters than that allowed under the territorial system oftaxation sanctioned by the GATT council understanding; and (3) the understanding adopted

by GATT did not optimally promote GATT policies

10 S Report 2708, 97th Cong., 2d Sess., 128 Cong Rec S7895 (1982); the bill died

13 The ETI Act created three new Code Sections, 941, 942, and 943, under Subpart

E of the Internal Revenue Code Section 941 defines qualifying foreign trade income withrespect to any transaction as the amount of gross income which, if excluded, will result

in a reduction of the taxable income of the taxpayer from such transaction equal to thegreater of: (1) 30% of the foreign sale and leasing income derived by the taxpayer from thetransaction, or (2) 1.2% of the foreign trading gross receipts derived by the taxpayer fromthe transaction, or (3) 15% of the foreign trade income derived by the taxpayer from thetransaction The ETI Act also created Code Section 114, providing for an exclusion fromgross income of qualifying income

14 Tax Reform Act of 1986, P.L 99–514

15 The specific tax law changes considered include: (1) the changes in depreciationtax rules; (2) the repeal of the investment tax credit; and (3) the reduction in the corporatestatutory income tax rate

16 The Census export data are compiled from U.S Custom’s documents that reflect allshipments from U.S ports using the “free alongside ship” value of merchandise withoutregard to the profitability of the exporter See, for example,Bureau of the Census (2002a)

for a discussion of trade data sources Because the exporting firms change somewhat fromquarter to quarter, we do not attempt to specify specific firm types for each quarter In 2002,the Census Bureau identified 246,452 exporting companies with 219,390 small companies(less than 100 employees), 19,139 medium companies, and 7,923 large companies (morethan 500 employees) Although smaller in number, large companies represented 71% ofexport value (Bureau of the Census, 2002b)

17 Machine readable versions of these data and the other macroeconomic variables wereobtained fromCITIBASE (1978)as amended

18 The quarterly R&D amounts are the result of linear interpolation of the annual series,subject to the condition that their sum for each year should add up to the corresponding an-nual value This is accomplished following methods used inDiz (1966)andEl-Sharif (1979)

19 From an initial list of 20 product classes with varying levels of missing data over the1967–1998 period, we eliminated service-related product classes (not generally eligiblefor the DISC/FSC benefit) and other classes deemed to be too broadly defined

20 The model is estimated using a Gauss-Marquardt maximum likelihood method

to simultaneously solve for both the model’s parameter estimates and its autoregressiverelationship (Harvey, 1981; Judge et al., 1985) That is, the relationship between the current-and prior-period error terms are modeled concurrently with the structural parameters Thisapproach provides more accurate standard error estimates than an ordinary least squaresestimation

The Effect of Export Tax Incentives on Export Volume 25

21 An autoregressive process of up to five periods was modeled for each regression.Nonsignificant lags were removed from any model; most models were accurately specifiedwith a single period lag, AR(1)

22 Such regressions produce high R2 values, low Durbin-Watson statistics, and low

standard errors In addition, the t- and F-statistics are invalid because the error terms

are nonnormally distributed and may be autocorrelated Furthermore, if the time seriesanalyzed are cointegrated, ordinary regression estimates can be biased in small samples

As shown byStock (1987), least square estimators of cointegrating parameters in staticlevel regressions are superconsistent, theoretically converging to their probability limits atfaster rates than conventional values

23 SeeKwiatkowski et al (1992)for a complete description of this testing procedure

24 Using the yields on the 20- and the 30-year U.S Government bonds does not alterthe results with respect to the statistical significance of the parameter estimates

25 The intercept and slope parameter estimates at any given point in time are determined

by taking the base year estimates (␤0and␤1) and adjusting them for cumulative changes up

to that point For example, for the post-1977 period, the intercept and the slope estimates are(␤0+ ␤2+ ␤4) and (␤1+ ␤3+ ␤5), respectively For the post-1984 period, the interceptand the slope estimates are, respectively (␤0+ ␤2+ ␤4+ ␤6) and (␤1+ ␤3+ ␤5+ ␤7)

26 The results of these robustness tests are available on request from the authors

27 See alsoWorld Trade Organization (2002, Annex A) for alternative calculations ofthe export subsidy provided to certain product categories in 2000

28 The European-American Business Council (2001)produced a more recent studythat examines potential products that might be targeted in the EU FSC/ETI trade retaliation.Although their estimates are not necessarily based on product elasticities, the reportindicates that products such as aircraft parts, machinery, vehicles, and electrical equipmentaccount for a substantial amount of U.S exports to Europe These products are consistentwith product classes 1, 2, 3, and 9 in the current study and these classes demonstrated themost consistent sensitivity to the changes in export tax incentives

ACKNOWLEDGMENTS

The authors gratefully acknowledge the financial support of the Ernst & YoungFoundation through its Tax Research Grant Program The authors thank JongsubKim for able research assistance, Wei Li for his comments on the econometrics,and the two referees and editor for their helpful comments

REFERENCES

Auquier, A., & Caves, R (1979) Monopolistic export industries, trade taxes and optimal competition

policy Economic Journal, 89, 559–581.

Bello, C., & Williamson, N C (1985) The American export trading company: Designing a new

international marketing institution Journal of Marketing, 49, 60–69.

26 B ANTHONY BILLINGS, GARY A McGILL AND MBODJA MOUGOU ´E

Bernard, A B., & Jensen, J B (1998) Understanding the U.S export boom Working Paper 6438.

National Bureau of Economic Research.

Bilkey, W (1982) Variables associated with export profitability Journal of International Business

Studies, 13, 39–55.

Brander, J., & Spencer, B (1983a) Strategic commitment with R&D: The symmetric case Bell Journal

of Economics, 14, 225–235.

Brander, J., & Spencer, B (1983b) Tariffs and the extraction of foreign monopoly rents under potential

entry Canadian Journal of Economics, 14, 371–389.

Bureau of the Census, Foreign Trade Division (2002a) U.S international trade in goods and services, Report FT900 (CB-02-132) (August).

Bureau of the Census, Economics and Statistics Administration (2002b) Advance report: A profile of U.S exporting companies, 2000 (CB-01-136) (October 18).

Campbell, J C., Ito, K., Pucik, V., & Yu, C M (1987) The Japanese and the U.S tax systems:

Implications for Japanese auto exports MIR, 2, 32–45.

Chowdhury, A R (1993) Does exchange rate volatility depress trade flows? Evidence from error

correction models Reviewof Economics and Statistics, 75, 700–706.

CITIBASE (1978) Citibank Economic Database (machine-readable magnetic data file), 1946 New York: Citibank, N A.

Collie, D (1991) Export subsidies and countervailing tariffs Journal of International Economics, 31,

309–324.

Czinkota, M R., & Johnston, W J (1985) Exporting: Does sales volume make a difference? Reply.

Journal of International Business Studies, 16, 157–162.

Dickey, D A., & Fuller, W A (1981) Likelihood ratio statistics for autoregressive time series with a

unit root Econometrica, 49, 1057–1072.

Diz, A C (1966) Money and Prices in Argentina, 1935–1962 Unpublished doctoral dissertation, University of Chicago, Chicago.

Dollar, D (1987) Import quotas and the product cycle Quarterly Journal of Economics, 102, 615–632 Dutton, J (1990) Targeted export subsidies as an exercise of monopoly power Canadian Journal of

Dis-Worldwide Tax Daily, 2002 WTD 14–27.

Feenstra, R (1986) Trade policy with several goods and market linkages Journal of International

Economics, 20, 249–267.

Fuller, W A (1976) Introduction to statistical time series New York: Wiley.

Goldstein, M., & Mohsin, K (1987) The supply and demand for exports: A simultaneous approach.

Reviewof Economics and Statistics, 60, 275–285.

Granger, C W J., & Newbold, P (1974) Spurious regressions in econometrics Journal of

Economet-rics, 26, 1045–1066.

Gravelle, J., Hughes, K., & Farb, W E (1976) The domestic international sales corporation and its

effects on U.S foreign trade and unemployment Congressional Research Service Washington,

DC.

Green, R T., & Lutz, J M (1978) The role of technology in export trade The United States and World

Trade, 131–147.

Hartzok, J (1985) Domestic international sales corporation returns, 1980 Compendium of Studies of

International Income and Taxes Statistics of Income, Internal Revenue Service, pp 289–309.

The Effect of Export Tax Incentives on Export Volume 27

Harvey, A C (1981) The econometric analysis of time series New York: Wiley.

Itoh, M., & Kiyono, K (1987) Welfare-enhancing export subsidies Journal of Political Economy, 95,

115–137.

Jackson, J H (1978) The jurisprudence of international trade: The DISC case in GATT American

Journal of International Law, 72, 747–757.

Jacobs, F A., & Larkins, E R (1998) Export tax incentives for establishing foreign markets: An

analysis of marginal costing techniques Accounting Horizons, 12, 374–396.

Judge, G G., Griffiths, W E., Hill, R C., & Lee, T C (1985) The theory and practice of econometrics.

New York: Wiley.

Karlsson, C (1988) Innovation adoption and the product life cycle Umea Economic Studies Series

No 185.

Kim, S W., & Lyn, E O (1987) Foreign direct investment theories, entry barriers, and reverse

in-vestments in U.S manufacturing industries Journal of International Business Studies, 18,

53–66.

Koo, A Y C., & Martin, S (1984) Market structure and U.S trade flows International Journal of

Industrial Organization, 2, 173–198.

Kwiatkowski, D., Phillips, P C B., Schmidt, P., & Shin, Y (1992) Testing the null hypothesis of

stationarity against the alternative of a unit root Journal of Econometrics, 54, 159–178 Larkins, E R (1991) Alternative tax vehicles for exportation (Part II) The Tax Adviser, 22, 247–257 Laussel, D (1992) Strategic commercial policy revisited: A supply function equilibrium model Amer-

ican Economic Review, 82, 84–99.

Lee, H., & Stone, J A (1994) Product and process innovation in the product life cycle: Estimates for

U.S manufacturing industries Southern Economic Journal, 60, 754–763.

Lewis-Beck, M (1986) Interrupted time series In: W D Berry & M S Lewis-Beck (Eds), NewTools

for Social Sciences: Advances and Applications in Research Methods (pp 209–240) Beverly

Hills: Sage.

Mansfield, E., Romeo, A., & Wagner, S (1979) Foreign trade and U.S research and development.

Reviewof Economics and Statistics, 61, 49–57.

National Science Foundation (various) Research and development in industry Surveys of Science

Resources Series.

Nolle, D (1991) An empirical analysis of market structure and import and export performance for

U.S manufacturing industries Quarterly Reviewof Economics and Business, 31, 59–78 Ohno, K (1990) Exchange rate fluctuations, pass-through, and market share International Monetary

Fund Staff Papers, 37, 294–310.

O’Neill, H M., & Ross, W (1991) Exchange rate and income effects on South Korean exports: The

U.S case Journal of Economic Development, 16, 87–111.

Phillips, P C B (1986) Understanding spurious regressions in econometrics Journal of Econometrics,

Rosson, P J., & Ford, D L (1982) Manufacturers overseas distributor relations and export

perfor-mance Journal of International Business Studies, 13, 57–72.

Rousslang, D J (1987) The effects of recent corporate tax changes on international trade National

Tax Journal, 40, 603–615.

Schneeweis, T (1985) A note on international trade and market structure Journal of International

Business Studies, 16, 139–152.

28 B ANTHONY BILLINGS, GARY A McGILL AND MBODJA MOUGOU ´E

Stadler, M (1991) R&D dynamics in the product life cycle Journal of Evolutionary Economics, 1,

293–305.

Stock, J H (1987) Asymptotic properties of least squares estimators of cointegrating vectors

Econo-metrica, 55, 1035–1056.

U.S Department of Commerce (various, 1967–1998) Survey of Current Business 47–78 Bureau of

Economic Analysis, U.S Department of Commerce, Washington, DC.

U.S Department of Treasury (1978) The Operation and Effect of the Domestic International Sales Corporation Legislation: 1976 Annual Report (April).

U.S Department of Treasury (1983) The operation and effect of the DISC: 1981 Annual Report U.S Department of Treasury (1993) The operation and effect of the foreign sales corporation legislation (January).

Vernon, R (1966) International investment and international trade in the product cycle Quarterly

Journal of Economics, 80, 190–207.

Wells, L T (1969) Test of a product cycle model of international trade: U.S exports of consumer

durables Quarterly Journal of Economics, 83, 152–162.

World Trade Law (1977) GATT: DISC and other discriminating income taxes Journal of World Trade

Law, 564–566.

World Trade Organization (2000) United States – Tax treatment for “foreign sales corporations”, Report of the Appellate Body; AB-1999-9 (February 24).

World Trade Organization (2002) United States – Tax treatment for “foreign sales corporations”,

recourse to arbitration by the United States under article 22.6 of the DSU and article 4.11 of the SCM agreement (August 30).

Yang, J (1996) Exchange rate changes and pricing behavior of U.S exporters Reviewof International

Economics, 4, 339–354.

IMPLICATIONS OF BENCHMARK

STATE AND LOCAL TAX RATES FOR MEASURES OF ESTIMATED IMPLICIT TAXES

by implicit tax theory This conclusion regarding the implicit tax hypothesis may be premature whenever the incidence of state and local income taxes contributes to this empirical finding First, state income taxes, treated as a negative subsidy when the benchmark state and local tax rate is set at zero, will likely cause implicit taxes to be underestimated Second, the observed relationship between estimated implicit taxes and pretax returns appears

to be reversible depending upon the researcher’s election of a statutory tax rate that incorporates the selected benchmark state and local tax rate The present study uses a sample of 848 firms covering the years from 1989 through 1998 to show how the relation between estimated implicit taxes and pretax returns can be manipulated by the selection of the benchmark

30 BRADLEY D CHILDS

state and local tax rate Since choosing an accurate benchmark state and local tax rate can be problematic, the present study suggests adjusting both estimated implicit taxes and pretax income by the amount of state and local income taxes incurred The results, using the regression model making this adjustment, appear to nullify the negative bias of a zero tax rate as the benchmark state and local tax rate.

INTRODUCTION

a benchmark asset that can be used to compare pretax returns This approach doesnot emphasize a benchmark ordinary income tax rate for the calculation of implicittaxes, presumably because it should not matter In theory, firms located in high-tax(low-tax) jurisdictions should have higher (lower) pretax profitability, whichshould render meaningless the selection of a tax rate benchmark Perhaps for thisreason, prior researchers (Callihan & White, 1999; Wilkie, 1992) have implicitlyassumed benchmark state and local tax rates to be zero.1 One rationale for azero benchmark rate is the idea that if a firm creates nexus with a state that has acorporate income tax, then operations in that state will produce enough additionalbenefits (i.e implicit tax benefits) to offset the incurrence of state income taxes

In the real world, these offsetting benefits may not arise, and even if they

do, measurement issues surrounding implicit taxes in the corporate sector mayprovide no evidence in support of the implicit tax hypothesis.2The current studypresents evidence that indicates that the choice of a benchmark state and local taxrate is an important decision to be made by empirical researchers In the imperfectworld of empirical analysis of financial statement data, a low benchmark stateand local tax rate may lead to a negative bias in the relationship between pretaxprofitability and estimated implicit taxes This potential bias can lead researchersand their audience to prematurely conclude that there is evidence of implicit taxes

in the corporate sector.3

One way to control for this potential bias is to select the proper benchmark, butthis is a daunting task given the lack of multijurisdictional data in the financialstatements of public companies Accordingly, the current study controls for thispotential bias by subtracting current state and local income tax expenses fromcomputations of pretax income and estimated implicit taxes After removingthe potential negative bias associated with a low benchmark rate for state andlocal income taxes, the results do not support implicit tax theory While theseresults suggest non-trivial market frictions, there are several other factors (e.g.risk differences, accounting method differences and industry differences) that

Implications of Benchmark State and Local Tax Rates 31may account for these results It is beyond the scope of this article to tackle thesedifficult empirical issues, but future research will need to control for some ofthese differences if the expected relation between pretax profitability and implicittaxes estimated from tax subsidies is to be observed.

The remainder of this article is organized as follows First, implicit tax theory

is reviewed and the effects of omitting state and local tax rates is developed.Second, the empirical procedures are discussed Finally, the descriptive statisticsand results are analyzed and conclusions are made

IMPLICIT TAX THEORY

Using notation fromScholes et al (2002, pp 98–99), let Rbbe the risk-adjusted

before-tax return on the benchmark investment Let Ra be the risk-adjusted

before-tax return on a particular investment, and let r∗be the common after-taxreturn that would occur in a perfectly competitive equilibrium

Rb− Ra= Implicit tax for the alternative investment. (1)Using the preceding definitions andEq (1), the following equation can be derived:

r∗= Rb(1− t) = Ra(1− t) +TSa

where,

t: the statutory tax rate,

TSa: the tax subsidy or savings generated by the alternative investment, and

Ia: the alternative investment

With further substitution,Eq (2)becomes

PTIa,b: the pretax income generated from the respective investments, and

Ib: the benchmark investment

The definition of TSa, the tax subsidy of an alternative investment is as follows:

32 BRADLEY D CHILDSwhere,

CTEa: current tax expense of the alternative investment

have been given pretax income vs the actual tax liability UsingEqs (3) and (4),

an estimated implicit tax for a firm can be calculated as follows:

IMPLICITi = ((PTIi )t− CTEi)/(1 − t)

I i

(5)where

IMPLICITi: the estimated implicit tax for firm i,

PTIi: pretax income for firm i,

CTEi: all current income taxes for firm i, and

I i: the investment for firm i.

The estimation of implicit taxes as presented inEq (5) is the method that hasbeen used byWilkie (1992)andCallihan and White (1999)in their exploration

of implicit taxes in the corporate sector Both studies found a negative relationbetween estimated implicit taxes and pretax returns, althoughCallihan and White

were significantly associated with estimated implicit taxes.4Their interpretation

of reduced implicit taxes for firms benefiting from market structure has beencriticized on methodological grounds byWright (2001)

In the prior research, the benchmark tax rate was limited to the federal tax rateonly, but most domestic firms are subject to state and local income taxes or theirequivalents.5A benchmark tax rate that excludes state and local income tax rateswill lead to underestimating tax subsidies as follow:

UnderTSa= [(PTIa)d− CTEa]− [(PTIa)f− CTEa] (6)where,

UnderTSa: the underestimated tax subsidy for the alternative investment,

d: the domestic statutory rate or [f + s(1 − f)],

s: the state and local statutory rate, and

f: the federal statutory rate

of the federal tax benefit of the deductibility of state and local income taxes